DLS.py

'''
Title : Assignment 2 - Duckworth Lewis Method
Name : Adarsh Shah, IISC MTech(AI)
SR no : 19473
'''
import numpy as np
import pandas as pd
from matplotlib import pyplot
from math import exp,sqrt
from scipy import optimize as op

# Data Preprocessing

data = pd.read_csv('./data/04_cricket_1999to2011.csv')
data = data[data['Innings']==1]
for i in range(len(data)):
    if data.iloc[i]['Over']==1:
        data.iloc[i]['Total.Runs'] = data.iloc[i]['Runs']
    else:
        data.iloc[i]['Total.Runs'] = data.iloc[i]['Runs'] + data.iloc[i-1]['Total.Runs']
country_data = data.filter(items=['Over','Innings.Total.Runs','Total.Runs','Wickets.in.Hand'])


# Objective function 1

def obj1(x,overs,runs):
    z,b = x
    func = np.multiply(z,np.subtract(1,np.exp(np.multiply(-1*b,overs))))
    error = np.subtract(runs,func)
    error = sqrt(np.sum(np.square(error)))/len(runs)
    return error

def DuckworthLewis20Params():
    Z0 = np.zeros(10)
    b = np.zeros(10)
    errors = np.zeros(10)
    cdata = country_data
    for w in range(1,11):
        d = cdata[cdata['Wickets.in.Hand']==w]
        overs = 50-d['Over']
        runs = d['Innings.Total.Runs']-d['Total.Runs']
        Z0[w-1],b[w-1]=op.fmin(func=obj1,x0=np.array([250,0.1]),args=(overs,runs),disp=False)
        errors[w-1] = obj1([Z0[w-1],b[w-1]],overs,runs)
    return Z0,b,errors

def plot1(Z,b):
    func = lambda w,o : Z[w-1]*(1-exp(-1*b[w-1]*o))
    for w in range(1,11):
        x = range(0,51)
        y = np.array([func(w,o) for o in range(0,51)])
        pyplot.plot(x,y,label=str(w))
        pyplot.text(x[-5],y[-5],w)
    pyplot.show()

# Test
Z0,b,errors = DuckworthLewis20Params()
plot1(Z0,b)
print('Z0\t\tb\t\terrors')
for i in range(10):
    print(f'{Z0[i]:.2f}\t\t{b[i]:.5f}\t\t{errors[i]:.2f}')


# Objective Function 2

def obj2(Z,data):
    L=Z[10]
    error=0
    for w in range(1,11):
        d = data[data['Wickets.in.Hand']==w]
        overs = 50-d['Over']
        runs = d['Innings.Total.Runs']-d['Total.Runs']
        fruns = Z[w-1]*np.subtract(1,np.exp(np.multiply(-1*L/Z[w-1],overs)))
        error+=sqrt(np.sum(np.square(np.subtract(runs,fruns))))/len(data)
    return error

def DuckworthLewis11Params():
    x0 = 250*np.ones(11)
    x0[10]=0.1
    cdata = country_data
    errors = np.zeros(10)
    res=op.minimize(fun=obj2,x0=x0,args=(cdata))
    Z = res.x[:10]
    L = res.x[10]
    for w in range(1,11):
        d = cdata[cdata['Wickets.in.Hand']==w]
        overs = 50-d['Over']
        runs = d['Innings.Total.Runs']-d['Total.Runs']
        fruns = Z[w-1]*np.subtract(1,np.exp(np.multiply(-1*L/Z[w-1],overs)))
        errors[w-1]=sqrt(np.sum(np.square(np.subtract(runs,fruns))))/len(cdata)
    return Z,L,errors

def plot2(Z,L):
    func = lambda w,o : Z[w-1]*(1-exp(-1*L*o/Z[w-1]))
    for w in range(1,11):
        x = range(0,51)
        y = np.array([func(w,o) for o in range(0,51)])
        pyplot.plot(x,y,label=str(w))
        pyplot.text(x[-5],y[-5],w)
    pyplot.show()


# Test
Z0,L,errors = DuckworthLewis11Params()
plot2(Z0,L)
print('Z0\t\tL\t\terrors')
for i in range(10):
    print(f'{Z0[i]:.2f}\t\t{L:.5f}\t\t{errors[i]:.2f}')